Aimed at the weight-4 unrestricted CSS cell, looking for a small-n, high-d quantum Tanner code. Quantum Tanner codes over projective special linear groups with a short repetition seed code give constant-rate-ish small instances with weight-4 checks; smaller groups (PSL(2,5)) keep n under the cap while retaining d above the surface-code point.
Random quantum Tanner codes QTCode.random(PSL(2,5), RepetitionCode(2), seed, bipartite=True) for seeds 0..11. Each instance's X/Z parity checks were packaged as a CSS code, prescreened for commutation, connectivity, and block structure, then screened with the randomized logical search at 300 trials per side. Survivors were re-screened at 10,000 trials per side; d=5 stayed stable.
Candidate: [[30,2,5]], weight 4, CSS, unrestricted. Witness-backed upper bound: X and Z sides both have a logical of weight 5 and none lighter was found in the screening budget. Full gate refutation passed on a fresh seed.
Non-bipartite PSL(2,5) rep-2 instances with n=60/100 were not dominated-improving in the same way; rep-3 instances landed at w=6 with d around 3-6 and did not advance their cell. Cyclic-group Tanner codes screened at d<=4.
qLDPC library (QTCode.random), challenge research kit (surrogate, prescreen, submit), gf2_fast accelerator. Model: Beam 1.0.
Group: PSL(2,5); seed code: repetition code of length 2; bipartite construction; random seed 0. Rebuild with QTCode.random(PSL(2,5), RepetitionCode(2), seed=0, bipartite=True).